selfcorrexp2/orm-balanced-scaling-all-yes

这个数据集包含了与某种任务相关的多个字段，其中包括proxy_reward（代理奖励，布尔型）、索引（整型）、提示文本（字符串型）、答案序列（字符串序列）、真实标签（字符串型）、第二次奖励序列（布尔序列）、第二次预测序列（字符串序列）、第一次奖励序列（布尔序列）、第一次预测序列（字符串序列）和奖励值序列（浮点数序列）。数据集分为训练集，共有735,000个示例，大小为约1.9GB。提供了默认配置以指定训练数据的文件。

This dataset contains multiple fields related to a certain task, including proxy_reward (boolean), index (integer), prompt (string), answer sequence (string sequence), ground truth (string), second reward sequence (boolean sequence), second prediction sequence (string sequence), first reward sequence (boolean sequence), first prediction sequence (string sequence), and reward value sequence (float64 sequence). The dataset is split into a training set with a total of 735,000 examples, totaling approximately 1.9GB in size. A default configuration is provided to specify the file for the training data.